[Scythels]

 

Where do you see the focus shift? I'm not sure I can. At least for my specific use it wouldn't be a problem even if there is some, although for an AF lens it would be most undesirable.

The depth of field primarily extends backwards as the lens is stopped down, so there is minor backward focal shift, as the front depth of field doesn't expand much.

 

 

As for the design, were you following them when they were designing the 14mm?

I wasn't following the design side of things when they were designing the 14mm, at least not with photographic lenses anyway.  I certainly feel well enough educated about that lens to comment on the production design, but not the process or any previous designs. 

 

 

I'd think they'd design optically for Nikon F mount, since their mount is slightly further out than Canon. As such they could easily extend the lens body for others. Unless the optics can stick out much further than the lens mount for whatever reason.

Designing "for a mount" is fairly relative, especially with a telephoto lens such as this.  The parameter for the distance from the vertex of the rear element to the image plane in Code V (optical CAD) is "IMC" or Image Clearance.  With a telephoto lens, IMC will almost definitely be >44 naturally, so there is no need to 'design' for any one particular long flange mount as it will all work out the same anyway.  Wide-angle retrofocus lenses which want to smash themselves up against the image plane in optimization must be designed for very specific IMC constraints (i.e mount specific) but telephoto lenses generally do not.

 

 

A 135mm f/2 lens would typically cover full frame sensor

Statements like these are quite dangerous.  Here: http://i.gyazo.com/d3247be61fadc9db9268390d769d6fb0.png is a working design for a camera lens.  The focal length of a lens is arbitrary for this design form (retrofocus) since a longer focal length will just mean a bigger lens for a bigger sensor with the same angular coverage.  The red bundle is the marginal ray - it traces from the optical axis to the center of the image plane and defines the f number.  The green bundle is the chief ray - it traces from the edge of the angular field to the edge of the image plane and defines the size sensor a lens covers as well as the field of view (the two are directly linked if the focal length is held constant).  You can see that shortly after the lens begins to strongly focus the light into a smaller bundle, the chief ray extends "further out" away from the marginal ray.  If you shrink the diameter of the lens elements, you will vignette the chief ray bundle there.  Ultimately, the size of the "back" elements determines the size sensor the lens covers.  It cannot be made arbitrarily large, typically the focusing within the lens limits the rear diameter to be +/- the front diameter but some designs are larger in the rear, though that makes for an odd design.  Because long focal length lenses have.. long focal lengths and large elements in front to allow their aperture (e.g f/2) their tendency is to "want" to cover a very large sensor.  However if there is vignetting that can be observed on a particular sensor size, one may be assured the lens does not 'cover' a much larger image area.  As this lens is a retrofocus lens, it could be extended to a sensor 2-3x larger (look at the angle of the chief ray coming out of the aperture stop) but everything after the front two elements would have to be very considerably enlarged.

 

 

I think the argument wasn't that it couldn't be made smaller/lighter, but it was to question how much smaller/lighter it could be

The volume and weight can be considerably reduced, but without a radical redesign the length must remain.  A lens barrel that tapers back to the lens mount without being very wide (e.g 100mm diameter of superteles) looks funny so it is often not done. There are most likely structural issues to that as well, but I am not a MechE.  In telephoto design the would-be-called-telephoto group i.e the first several elements dominates overall size and weight due to general practices in mechanical design.

 

 

A similar question often pops up on various forums, why aren't there longer focal length lenses (say, > 300mm) made specifically for APS-C sensors? The argument against it typically runs along the lines of the saving being insufficient to justify having two versions of the lens, so just make the one with the bigger image circle to cover both uses.

In a nutshell, in very long focal length situations (mirror lenses aside) the front elements are generally very large and determine the general diameter of the barrel.  The elements that could be reduced in size to cut off the image circle contribute little to the overall weight of the lens, and there isn't much sense in making a backwards-tapering lens barrel.  I would have shown a telephoto ray trace above but my student copy of Code V hasn't been approved for this semester yet and I don't love photozone enough to travel to the lab and back to produce a lens diagram for the forum